AXON: An Automated Netlist Optimization Framework for High-Speed Adders
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Adders are fundamental building blocks in modern digital systems, such as CPUs and accelerators, and their performance, power, and area (PPA) directly affect overall system efficiency. Contemporary adders commonly adopt parallelprefix architectures with well-known PPA trade-offs. However, for specific design objectives, these classical architectures often fail to achieve globally optimal PPA. Prior work often optimizes topology without netlist-/cell-level awareness, and generalpurpose synthesis heuristics are not adder-specific, leading to suboptimal PPA. To address these limitations, we propose an automated netlist optimization framework for adders, named AXON. The framework performs comprehensive design space exploration from the architectural to the netlist level, integrating prefix topology search with standard-cell-aware mapping strategies. It employs a hierarchical search methodology that rapidly converges to near-optimal solutions under given PPA constraints. Furthermore, we introduce a hybrid ultra-high-speed adder structure that combines the advantages of the parallelprefix and Ling architectures to effectively shorten the critical path. Experimental results based on the TSMC 28nm library show that AXON achieves up to 10.3%, 12.6%, and 32.1% improvement in delay, area-delay product, and energy-delay product, respectively, compared to the synthesis results using commercial synthesis tools.
Fields of Study & Subjects
VLSI and FPGA Design Techniques
Low-power high-performance VLSI design
Embedded Systems Design Techniques
Publication details
- Journal / source
- International Symposium of Electronics Design Automation (ISEDA)
- Pages
- 266–271
- Publication date
- 1 May 2026
- Language
- EN